diff --git "a/meta/camera.jsonl" "b/meta/camera.jsonl"
--- "a/meta/camera.jsonl"
+++ "b/meta/camera.jsonl"
@@ -707,3 +707,8 @@
 {"episode_index": 706, "cameras": [{"name": "opst_cam", "source_role": "front", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "side_cam", "source_role": "side", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "wrist_cam", "source_role": "wrist", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"serial": "218622278247"}}]}
 {"episode_index": 707, "cameras": [{"name": "opst_cam", "source_role": "front", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "side_cam", "source_role": "side", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "wrist_cam", "source_role": "wrist", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"serial": "218622278247"}}]}
 {"episode_index": 708, "cameras": [{"name": "opst_cam", "source_role": "front", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "side_cam", "source_role": "side", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "wrist_cam", "source_role": "wrist", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"serial": "218622278247"}}]}
+{"episode_index": 709, "cameras": [{"name": "opst_cam", "source_role": "front", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "side_cam", "source_role": "side", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "wrist_cam", "source_role": "wrist", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"serial": "218622278247"}}]}
+{"episode_index": 710, "cameras": [{"name": "opst_cam", "source_role": "front", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "side_cam", "source_role": "side", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "wrist_cam", "source_role": "wrist", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"serial": "218622278247"}}]}
+{"episode_index": 711, "cameras": [{"name": "opst_cam", "source_role": "front", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "side_cam", "source_role": "side", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "wrist_cam", "source_role": "wrist", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"serial": "218622278247"}}]}
+{"episode_index": 712, "cameras": [{"name": "opst_cam", "source_role": "front", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "side_cam", "source_role": "side", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "wrist_cam", "source_role": "wrist", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"serial": "218622278247"}}]}
+{"episode_index": 713, "cameras": [{"name": "opst_cam", "source_role": "front", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "side_cam", "source_role": "side", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"T_front_side": [[-0.011371483630178314, 0.051428717795708956, -0.9986119247965837, 0.40760295854929707], [-0.7843157594220633, 0.6190179879925587, 0.04081078367094121, 0.1025990384934167], [0.6202575907493924, 0.7836911493232097, 0.03329720093680111, 0.7251040190673153], [0.0, 0.0, 0.0, 1.0]], "fit_indices": [2, 3, 4, 5, 6], "heldout_indices": [7, 8], "fit_ray_rmse_equivalent_px_per_axis": 0.22539347598332427, "heldout_checks": [{"index": 7, "side_ray_rmse_equivalent_px": 0.4784647121129936, "relative_translation_difference_m": 0.0023584630267821914, "relative_rotation_difference_deg": 0.37122318782681246}, {"index": 8, "side_ray_rmse_equivalent_px": 1.3694367043740425, "relative_translation_difference_m": 0.003710327133054494, "relative_rotation_difference_deg": 0.37687740416111115}], "relative_calibration_validated": true, "solver_success": true, "schema": "front_side_rig_v1", "front_serial": "234222300865", "side_serial": "218722271409", "fixed_cameras": true, "output_frame": "right_base_estimate", "base_calibration_validated": false, "calibration_validated": false, "motion_enabled": false, "board": {"squares": [6, 6], "square_m": 0.05, "marker_m_provisional": 0.0375, "dictionary": "DICT_4X4_50"}, "position_semantics": "visible_surface_proxy; not a volumetric object centroid", "timestamp_note": "Global-time timestamp pairing; exposures are not hardware synchronized.", "sources": ["pose_000.json", "pose_001.json", "pose_002.json", "pose_003.json", "pose_004.json", "pose_005.json", "pose_006.json", "pose_007.json", "pose_008.json"], "excluded_indices": [0, 1], "exclusion_reason": "Side camera rotated between pose 1 and pose 3: static-background feature matches show median shift -17.28 px in side, 0 px in front; see background-pose1-to3-check.json. Retain the subsequent fixed-camera segment, poses 3-9; reserve poses 8-9 for held-out checking. Full-set rejected fit preserved in rig-all-poses-rejected.json.", "camera_profiles": {"front": {"camera_serial": "234222300865", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 387.2059631347656, "fy": 386.67889404296875, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "side": {"camera_serial": "218722271409", "pyrealsense2_version": "2.59.1.11053", "intrinsics": {"width": 640, "height": 480, "fx": 393.065185546875, "fy": 392.5256042480469, "ppx": 323.5518798828125, "ppy": 243.4866485595703, "coeffs": [-0.05449247732758522, 0.06090652197599411, 0.00045267402310855687, 0.0005052064079791307, -0.020919060334563255]}, "distortion_model": "distortion.inverse_brown_conrady"}}, "T_base_front_candidate": [[-0.009052521174897759, 0.79075214916911, -0.6120695143893488, 0.8604397321514018], [0.9989787735231245, -0.019944396740285957, -0.04054172034954435, 0.04282932213494495], [-0.044265809725030476, -0.6118114575774964, -0.7897640650639202, 0.7453855280919021], [0.0, 0.0, 0.0, 1.0]], "T_base_side_candidate": [[-0.999737194160135, 0.009350783625123016, 0.020930968883886437, 0.49406624309724007], [-0.020863475883285503, 0.0072580696692524935, -0.9997559881284753, 0.41857278545667637], [-0.009500420353315076, -0.9999299391808775, -0.007061072357504628, 0.09191028810590329], [0.0, 0.0, 0.0, 1.0]], "base_reference_sha256": "49f9e67ce7e9a3b3828b884bcd7de460ed7af7b07c73df97de8af7a356e65922", "operator_confirmed_front_and_base_unchanged": true, "depth_corrections": {"front": {"schema": "krvla_rgb_depth_correction_v1", "camera_serial": "234222300865", "formula": "corrected_z_m = 1 / (1 / raw_z_m + inverse_depth_offset_per_m)", "inverse_depth_offset_per_m": 0.0747559004574122, "raw_depth_domain_m": [0.7990000379504636, 1.0330000490648672], "profile": {"intrinsics": {"width": 640, "height": 480, "fx": 387.4169616699219, "fy": 386.8896179199219, "ppx": 324.240966796875, "ppy": 240.64044189453125, "coeffs": [-0.055842526257038116, 0.06478709727525711, -0.0005568224587477744, 0.0009001302532851696, -0.02112000063061714]}, "distortion_model": "distortion.inverse_brown_conrady"}, "reference": "User-confirmed 310 x 310 mm, 10 x 10 square board; recorded RGB PnP poses assumed correct", "status": "empirical_rgb_referenced_depth_correction", "base_calibration_validated": false, "scene_generalization_validated": false, "outside_domain_policy": "Extrapolate the fitted inverse-depth model within the existing front tracking depth range [0.1,1.5] m; flag estimates outside the measured [0.799,1.033] m support. Zero invalid/out-of-application-range pixels.", "scope": "Recorded board workspace and depth range; target-independent accuracy not established", "source_fit_archive_sha256": "3e4688a58066e2f6ea2ee3aeadd2bab58d2779b96fe7ff8293dafcad2a9ab1b3", "application_depth_range_m": [0.1, 1.5], "application_scope_note": "Extension for the recorded manipulation workspace; extrapolated accuracy is not independently established."}}, "base_transform_acceptance": {"schema": "krvla_operator_transform_acceptance_v1", "operator_accepted": true, "created_utc": "2026-11-07T23:06:45.193593+00:00", "transform_sha256": "b68ce375b4e05c99110cf8e4ccbc9474f045e5872d45515918f84f20b682e3be", "source_rig": "/home/ailab/Desktop/Kinematic-Residual-VLA/scratch/real_robot/front-side-calibration-20261020-01/rig.json", "source_rig_sha256": "55b0d0b294aafedf086dbfa4005547cb156a3ba420b2227de72a1091eb370a07", "source_handeye": "scratch/real_robot/right-handeye/session-20261019-03/candidate-calibration.json", "operator_instruction": "i think i had already done the camera to robot transformation long ago. That is literally how we were tracking the object velocities. Lets use that", "mounts_unchanged_operator_confirmed": true, "base_calibration_validated": false, "existing_verification": {"poses": 5, "position_mean_m": 0.007242830108009557, "position_max_m": 0.008717014332750316, "orientation_mean_deg": 0.5984637253406115, "orientation_max_deg": 0.8404472274041179, "proposed_position_threshold_m": 0.005, "meets_proposed_position_threshold": false, "absolute_base_frame_accuracy_measured": false, "board_length_user_confirmed_m": 0.3, "next_check": "Independent physical point/TCP check in the right base frame; further repeated images alone do not establish absolute frame accuracy."}, "scope": "Use the existing transform for KRVLA training and collection; does not assert independent validation or change depth processing."}}}, {"name": "wrist_cam", "source_role": "wrist", "width": 384, "height": 384, "image_geometry": {"crop_xywh": [0, 0, 640, 480], "front_quad_xy": null, "rectified_size_wh": [640, 480], "homography_raw_to_rectified": [[1.0, 0.0, 0.0], [0.0, 1.0, 0.0], [0.0, 0.0, 1.0]], "resized_size_wh": [384, 288], "padding_ltrb": [0, 96, 0, 0], "pixel_transform_raw_to_exported": [[0.6, 0.0, -0.2], [0.0, 0.6, 95.8], [0.0, 0.0, 1.0]]}, "calibration_image_space": "original raw 640x480; apply image_geometry pixel transform", "calibration": {"serial": "218622278247"}}]}